python队列通信：rabbitMQ使用

（一）、前言
为什么引入消息队列？
1.程序解耦
2.提升性能
3.降低多业务逻辑复杂度

（二）、python操作rabbit mq
rabbitmq配置安装基本使用参见上节文章，不再复述。
若想使用python操作rabbitmq，需安装pika模块，直接pip安装：
pip install pika

1.最简单的rabbitmq producer端与consumer端对话：
producer：

#Author :ywqimport pikaauth=pika.PlainCredentials('ywq','qwe')  #save auth indoconnection = pika.BlockingConnection(pika.ConnectionParameters(               '192.168.0.158',5672,'/',auth))  #connect to rabbitchannel = connection.channel()   #create channelchannel.queue_declare(queue='hello')   #declare queue#n RabbitMQ a message can never be sent directly to the queue, it always needs to go through an exchange.channel.basic_publish(exchange='',                      routing_key='hello',                      body='Hello World!')    #the body is the msg contentprint(" [x] Sent 'Hello World!'")connection.close()

consumer：

#Author :ywqimport pikaauth=pika.PlainCredentials('ywq','qwe') #auth infoconnection = pika.BlockingConnection(pika.ConnectionParameters(               '192.168.0.158',5672,'/',auth))   #connect to rabbitchannel = connection.channel()          #create channelchannel.queue_declare(queue='hello')      #decalre queuedef callback(ch, method, properties, body):    print(" [x] Received %r" % body)channel.basic_consume(callback,                      queue='hello',                      no_ack=True)print(' [*] Waiting for messages. To exit press CTRL+C')channel.start_consuming()

消息传递消费过程中，可以在rabbit web管理页面实时查看队列消息信息。

2.持久化的消息队列，避免宕机等意外情况造成消息队列丢失。
consumer端无需改变，在producer端代码内加上两个属性，分别使消息持久化、队列持久化，只选其一还是会出现消息丢失，必须同时开启：

delivery_mode=2  #make msg persisdentdurable=True

属性插入位置见如下代码（producer端）：

#Author :ywqimport pika,sysauth_info=pika.PlainCredentials('ywq','qwe')connection=pika.BlockingConnection(pika.ConnectionParameters(             '192.168.0.158',5672,'/',auth_info        ))channel=connection.channel()channel.queue_declare(queue='test1',durable=True)  #durable=Ture, make queue persistentmsg=''.join(sys.argv[1:]) or 'Hello'channel.basic_publish(    exchange='',    routing_key='test1',    body=msg,    properties=pika.BasicProperties(            delivery_mode=2  #make msg persisdent    ))print('Send done:',msg)connection.close()

3.公平分发
在多consumer的情况下，默认rabbit是轮询发送消息的，但有的consumer消费速度快，有的消费速度慢，为了资源使用更平衡，引入ack确认机制。consumer消费完消息后会给rabbit发送ack，一旦未ack的消息数量超过指定允许的数量，则不再往该consumer发送，改为发送给其他consumer。

producer端代码不用改变，需要给consumer端代码插入两个属性：

channel.basic_qos(prefetch_count= *)   #define the max  non_ack_countchannel.basic_ack(delivery_tag=deliver.delivery_tag)  #send ack to rabbitmq

属性插入位置见如下代码（consumer端）：

#Author :ywqimport pika,timeauth_info=pika.PlainCredentials('ywq','qwe')connection=pika.BlockingConnection(pika.ConnectionParameters(    '192.168.0.158',5672,'/',auth_info    ))channel=connection.channel()channel.queue_declare(queue='test2',durable=True)def callback(chann,deliver,properties,body):    print('Recv:',body)    time.sleep(5)    chann.basic_ack(delivery_tag=deliver.delivery_tag)  #send ack to rabbitchannel.basic_qos(prefetch_count=1)'''注意，no_ack=False  注意，这里的no_ack类型仅仅是告诉rabbit该消费者队列是否返回ack，若要返回ack，需要在callback内定义prefetch_count=1,未ack的msg数量超过1个，则此consumer不再接受msg，此配置需写在channel.basic_consume上方，否则会造成non_ack情况出现。'''channel.basic_consume(    callback,    queue='test2')channel.start_consuming()

三、消息发布/订阅
上方的几种模式都是producer端发送一次，则consumer端接收一次，能不能实现一个producer发送，多个关联的consumer同时接收呢？of course，rabbit支持消息发布订阅，共支持三种模式，通过组件exchange转发器，实现3种模式：

fanout: 所有bind到此exchange的queue都可以接收消息，类似广播。
direct: 通过routingKey和exchange决定的哪个唯一的queue可以接收消息，推送给绑定了该queue的consumer，类似组播。
topic:所有符合routingKey(此时可以是一个表达式)的routingKey所bind的queue可以接收消息，类似前缀列表匹配路由。

1.fanout
publish端（producer）：

#Author :ywqimport pika,sys,timeauth_info=pika.PlainCredentials('ywq','qwe')connection=pika.BlockingConnection(pika.ConnectionParameters(    '192.168.0.158',5672,'/',auth_info    ))channel=connection.channel()channel.exchange_declare(exchange='hello',                         exchange_type='fanout'                         )msg=''.join(sys.argv[1:]) or 'Hello world %s' %time.time()channel.basic_publish(    exchange='hello',    routing_key='',    body=msg,    properties=pika.BasicProperties(        delivery_mode=2    ))print('send done')connection.close()

subscribe端（consumer）：

#Author :ywqimport pikaauth_info=pika.PlainCredentials('ywq','qwe')connection=pika.BlockingConnection(pika.ConnectionParameters(    '192.168.0.158',5672,'/',auth_info    ))channel=connection.channel()channel.exchange_declare(    exchange='hello',    exchange_type='fanout')random_num=channel.queue_declare(exclusive=True)  #随机与rabbit建立一个queue，comsumer断开后，该queue立即删除释放queue_name=random_num.method.queuechannel.basic_qos(prefetch_count=1)channel.queue_bind(    queue=queue_name,    exchange='hello')def callback(chann,deliver,properties,body):    print('Recv:',body)    chann.basic_ack(delivery_tag=deliver.delivery_tag)  #send ack to rabbitchannel.basic_consume(    callback,    queue=queue_name,)channel.start_consuming()

实现producer一次发送，多个关联consumer接收。
使用exchange模式时：
1.producer端不再申明queue，直接申明exchange
2.consumer端仍需绑定队列并指定exchange来接收message
3.consumer最好创建随机queue，使用完后立即释放。
随机队列名在web下可以检测到：

2.direct

使用exchange同时consumer有选择性的接收消息。队列绑定关键字，producer将数据根据关键字发送到消息exchange，exchange根据 关键字 判定应该将数据发送至指定队列，consumer相应接收。即在fanout基础上增加了routing key.

producer：

#Author :ywqimport pika,sysauth_info=pika.PlainCredentials('ywq','qwe')connection=pika.BlockingConnection(pika.ConnectionParameters(    '192.168.0.158',5672,'/',auth_info    ))channel=connection.channel()channel.exchange_declare(exchange='direct_log',                      exchange_type='direct',                      )while True:    route_key=input('Input  routing key:')    msg=''.join(sys.argv[1:]) or 'Hello'    channel.basic_publish(        exchange='direct_log',        routing_key=route_key,        body=msg,        properties=pika.BasicProperties(            delivery_mode=2        )    )connection.close()

consumer：

#Author :ywqimport pika,sysauth_info=pika.PlainCredentials('ywq','qwe')connection=pika.BlockingConnection(pika.ConnectionParameters(    '192.168.0.158',5672,'/',auth_info))channel=connection.channel()channel.exchange_declare(    exchange='direct_log',    exchange_type='direct')queue_num=channel.queue_declare(exclusive=True)queue_name=queue_num.method.queueroute_key=input('Input routing key:')channel.queue_bind(    queue=queue_name,    exchange='direct_log',    routing_key=route_key)def callback(chann,deliver,property,body):    print('Recv:[level:%s],[msg:%s]' %(route_key,body))    chann.basic_ack(delivery_tag=deliver.delivery_tag)channel.basic_qos(prefetch_count=1)channel.basic_consume(    callback,    queue=queue_name)channel.start_consuming()

同时开启多个consumer，其中两个接收notice，两个接收warning，运行效果如下：

3.topic

相较于direct，topic能实现模糊匹配式工作方式（在consumer端指定匹配方式），只要routing key包含指定的关键字，则将该msg发往绑定的queue上。

rabbitmq通配符规则：
符号“#”匹配一个或多个词，符号“”匹配一个词。因此“abc.#”能够匹配到“abc.m.n”，但是“abc.*‘’ 只会匹配到“abc.m”。‘.’号为分割符。使用通配符匹配时必须使用‘.’号分割。

producer：

#Author :ywqimport pika,sysauth_info=pika.PlainCredentials('ywq','qwe')connection=pika.BlockingConnection(pika.ConnectionParameters(    '192.168.0.158',5672,'/',auth_info    ))channel=connection.channel()channel.exchange_declare(exchange='topic_log',                      exchange_type='topic',                      )while True:    route_key=input('Input  routing key:')    msg=''.join(sys.argv[1:]) or 'Hello'    channel.basic_publish(        exchange='topic_log',        routing_key=route_key,        body=msg,        properties=pika.BasicProperties(            delivery_mode=2        )    )connection.close()

consumer：

#Author :ywqimport pika,sysauth_info=pika.PlainCredentials('ywq','qwe')connection=pika.BlockingConnection(pika.ConnectionParameters(    '192.168.0.158',5672,'/',auth_info))channel=connection.channel()channel.exchange_declare(    exchange='topic_log',    exchange_type='topic')queue_num=channel.queue_declare(exclusive=True)queue_name=queue_num.method.queueroute_key=input('Input routing key:')channel.queue_bind(    queue=queue_name,    exchange='topic_log',    routing_key=route_key)def callback(chann,deliver,property,body):    print('Recv:[type:%s],[msg:%s]' %(route_key,body))    chann.basic_ack(delivery_tag=deliver.delivery_tag)channel.basic_qos(prefetch_count=1)channel.basic_consume(    callback,    queue=queue_name)channel.start_consuming()

运行效果：

rabbitmq三种publish/subscribe模型简单介绍完毕，
以上！